Sampling for biological indicators of soil health: How does sampling methodology affect research results?

Abstract

A wide variety of soil sampling methods have been adopted, yet limited research has addressed how sampling methods can impact research results particularly for spatially variable soil properties. To address these concerns, we intensively sampled two existing soil health trials in Texas for analysis of soil extracellular enzymatic activities (EEAs), permanganate oxidizable C (POXC), soil organic C (SOC), and pH. The data generated was used to quantify the effect of composite sampling, sampling position (in-row vs between-row), and number of samples on statistical results. We compared a Monte Carlo simulation approach with a traditional power analysis to quantify the impact of the number of samples on our results. Calculated composite sampling resulted in lower statistically significant results compared to discrete sampling, demonstrating lower sensitivity of composite sampling. Sampling in-row resulted in greater values specifically for EEAs, pH, POXC, and SOC. Simulation generated estimated number of samples were mainly driven by the effect size and spatial variation. EEAs were highly variable, resulting in a higher estimated number of samples as compared to other soil properties such as POXC and SOC. In our study with two sites, soil with greater clay content resulted in overall lower estimated number of samples compared to a relatively sandy soil, demonstrating a greater sampling requirement for sandy soil. The simulation approach had a weak correlation with traditional power analysis. The simulation predicted lower estimated number of samples when compared to power analysis, which can help save cost in the long run. Overall, factors like composite sampling, sampling in-row vs between-row, and number of samples taken did impact research results. The impact of these sampling strategies on soil parameters of interest needs to be investigated to achieve the most accurate research results in soil science.